Executive Summary
Curity summarizes agentic AI incidents including prompt injection in GitHub's Agentic Workflows and the GitHub MCP server that could leak private repository data, and Salesforce Agentforce prompt injection.
What Happened
This post summarizes several agentic AI security issues, including prompt injection in GitHub’s Agentic Workflows and the GitHub MCP server that could leak private repository data. It also discusses Salesforce Agentforce prompt injection and other incidents involving untrusted content influencing agent behavior.
Affected Products
- GitHub Agentic Workflows
- GitHub MCP server
- Salesforce Agentforce
Vulnerability / Attack Technique
Prompt injection causing private data leaks
Technical Mechanism
Untrusted content such as issues or records influenced agents with access to private repositories or CRM data, leading to disclosure.
AI Security Category
LLM01:2025 Prompt Injection
Part of the Prompt Injection topic.
Part of the MCP Security topic.
Relevant Frameworks
RealGround Analysis
Agents connected to private data and public input are the recurring pattern. RealGround analysis: separate agents that read public content from those with private data access.
Recommended Controls
- Scope MCP server tokens to the repositories an agent needs
- Do not let agents act on public issues with private-repo access
- Apply vendor fixes for GitHub and Salesforce agent features
Related RealGround Service
Secure AI Agent Build · or start with the free AI Agent Security Scanner.
Sources
Curity: https://curity.io/blog/5-agentic-ai-security-incidents-and-vulnerabilities/
Summary and analysis written by RealGround from the source above; last updated 2026-10-17.